1. Field:
The present disclosure relates to methods and devices for improving image registration tools and using geomorphical algorithms to nominate areas that have a high probability of change between two georegistered images.
2. Background:
Methods for performing image processing have been used in many different applications, including agricultural and military applications. For example, an image may be taken over an area and automated methods used to identify features of interest in the image, such as buildings, automobiles, fields, or other objects. However, using a single image to automatically detect such objects may present difficulties in correctly assigning pixels to objects, leading to errors in the result of the automatic object detection process. A pixel may be a single point in an image.
One method to improve the result of the automatic change detection between images is to register two different images of the same physical location to each other, and then correlate corresponding pixels in the two images to enhance the probability that a given pixel of the combined image is correctly assigned to a given object. However, this method also has drawbacks. Existing change detection methods tend to attempt to isolate pixels that have changed, rather than areas of change. Correctly identifying areas of change would be more useful. Additionally, existing registration methods may contain many false positive change nominations, or detections. Many of the false positives would be a result of variations of any or all of the following: collection characteristics (altitude, obliquity, and others); temporal characteristics (sun angle, minor seasonal changes, and others); and post processing/display (stretch or resampling method).
Furthermore, when using known registering techniques, geospatial imagery is unnecessarily retaken regularly because it is not known whether a given image of an area has changed since a previous image of the area was taken. Retaking these images may be expensive.
Thus, new methods, techniques, and devices are desired to increase the accuracy of image processing and to avoid unnecessarily retaking geospatial images. Additional methods, techniques, and devices are desired to increase the accuracy of techniques for detecting changes between images, and to address the issues described above.